[unreadable] [unreadable] It is now widely recognized that there is a great need for more powerful automated methods to assist biomedical scientists in filtering, querying, and extracting information from the scientific literature. Building on our past research accomplishments in biomedical text mining, we plan to develop new algorithms and software systems that will significantly improve the ability of biomedical researchers to exploit the scientific literature. In particular, we plan to develop, evaluate and field systems that (1) aid in annotating high-throughput experiments by extracting and organizing information from text sources, and (2) assist genome database curators by identifying relevant articles and predicting appropriate ontology codes for specific query genes and proteins. In support of these systems, we plan to develop novel machine-learning based text-mining algorithms for training on coarsely labeled data, and inducing models of relationships among specific types of entities expressed in natural language. [unreadable] [unreadable] [unreadable]